Joined: 03 Jan 2010 Posts: 7568 Location: YWG, Canada
Posted: Sat Feb 17, 2018 8:29 Post subject:
tried no one seems to care. on top of that there is something wrong with ipq806x on k4.9+ that affects qos on top of that, adding latency. from what i see no one really knows, i see no fix attempts anywhere on any linux site or forum, i dont even see mentions of it besides my own posts
Joined: 30 Jan 2015 Posts: 676 Location: Texas, USA
Posted: Sat Mar 03, 2018 9:35 Post subject:
CAKE shaper looks impressive, would be great to have it dd-wrt.
https://www.bufferbloat.net/projects/attachments/150817135028_cake-battlemesh-v8.pdf _________________ ASUS GT-BE98 PRO Main: Fiber 5gbps up/down
ASUS AXE16000: AI Mesh node
2 X ASUS RT-AX89X: AI Mesh nodes
QNAP QSW-1208-8C 12-Port 10GbE Switch
XS712T ProSafe 12-Port 10GbE Switch
3 X R9000 DD-WRT Mesh
Joined: 03 Jan 2010 Posts: 7568 Location: YWG, Canada
Posted: Sat Mar 03, 2018 9:46 Post subject:
now to convince brainslayer...and convincing him of anything is extremely hard, and takes a long time. if anyone wants to bump the cake tickets.. _________________ LATEST FIRMWARE(S)
BrainSlayer wrote:
we just do it since we do not like any restrictions enforced by stupid cocaine snorting managers
Well it is easy to get impressive results if you compare it to a broken fq_codel. Thing is, only dd-wrt has a working fq_codel implementation at the moment, all others including openwrt ship fq_codel with wrong rate calculation, you have to thank tatsuya46 for constantly nagging as otherwise I wouldn't have found this. There was a change in early 3.x kernels that broken fq_codel and it has been broken since then, but not in dd-wrt:-) _________________ KONG PB's: http://www.desipro.de/ddwrt/
KONG Info: http://tips.desipro.de/
Joined: 30 Jan 2015 Posts: 676 Location: Texas, USA
Posted: Sat Mar 03, 2018 19:00 Post subject:
tatsuya46 wrote:
now to convince brainslayer...and convincing him of anything is extremely hard, and takes a long time. if anyone wants to bump the cake tickets..
I guess, BS is now fully into integrating Gateworks products into dd-wrt
I see in last few weeks no other major bugs fixed except the NTP fixes _________________ ASUS GT-BE98 PRO Main: Fiber 5gbps up/down
ASUS AXE16000: AI Mesh node
2 X ASUS RT-AX89X: AI Mesh nodes
QNAP QSW-1208-8C 12-Port 10GbE Switch
XS712T ProSafe 12-Port 10GbE Switch
3 X R9000 DD-WRT Mesh
Well it is easy to get impressive results if you compare it to a broken fq_codel. Thing is, only dd-wrt has a working fq_codel implementation at the moment, all others including openwrt ship fq_codel with wrong rate calculation, you have to thank tatsuya46 for constantly nagging as otherwise I wouldn't have found this. There was a change in early 3.x kernels that broken fq_codel and it has been broken since then, but not in dd-wrt:-)
@Kong, We all appreciative of the current QOS functionality as it is working, but there is more scope for improvement...
All we look for now is a take it up a notch approach _________________ ASUS GT-BE98 PRO Main: Fiber 5gbps up/down
ASUS AXE16000: AI Mesh node
2 X ASUS RT-AX89X: AI Mesh nodes
QNAP QSW-1208-8C 12-Port 10GbE Switch
XS712T ProSafe 12-Port 10GbE Switch
3 X R9000 DD-WRT Mesh
Well it is easy to get impressive results if you compare it to a broken fq_codel. Thing is, only dd-wrt has a working fq_codel implementation at the moment, all others including openwrt ship fq_codel with wrong rate calculation, you have to thank tatsuya46 for constantly nagging as otherwise I wouldn't have found this. There was a change in early 3.x kernels that broken fq_codel and it has been broken since then, but not in dd-wrt:-)
So, It seems the difference B/W fq_codel and cake won't be worth it in DD-WRT as opposed to OpenWRT/LEDE since they only saw better performance from cake because of the bad implementation of fq_codel from their side.
From what I've been reading, It seems the state of how well QOS works and what is broken and what depends on the kernel.
How is QOS in DD-WRT in Kernal 3.XX (My router = 841n v9) Is it majorly broken? _________________ Router: TP-Link TL-WR841N V9
Joined: 03 Jan 2010 Posts: 7568 Location: YWG, Canada
Posted: Mon Mar 05, 2018 17:51 Post subject:
k3.x is "fine" (its got its large list of issues but about the double latency i said, its free of that)
cake still at least looks better, & the reduced cpu usage it apparently has too should alone make it worth it.. its also said to be more fair when a connection starts spawning more and more connections _________________ LATEST FIRMWARE(S)
BrainSlayer wrote:
we just do it since we do not like any restrictions enforced by stupid cocaine snorting managers